1 Computer ASSisted ANalysis using Dynamic Regression AlgorithmsNow-casting and structural breaks analysis techniques for smart statistics Computer ASSisted ANalysis using Dynamic Regression Algorithms CASSANDRA JRC / EUROSTAT common activities on Big Data analytics 1

2 Cassandra goals: Exploit new "big" data tools to extract more value from official statistics Develop a platform for now-casting of infra- annual indicators of policy relevance Explore techniques to identify structural breaks in time series Add value towards "Smart statistics"

3 The challenge VVVs of data Evidence based policy making …

4 Nowcasting: estimate current value of indicator

5 SARIMA approach

6 Now-casting = forecasting r periods after TSCurrent solution Machine learning Limited "new" sources r between 1 and 6 Confidence interval depends on use Conf. interval r=1 t Time series

7 Cassandra: Now-castingFlexible, modular structure

8 Finding structural breaks = "forecasting" last TS valueCurrent solution Only "simple" model r is 1 Confidence interval depends on how sensitive Conf. interval r=1 t Time series (to t-1)

9 Real life application: structural break in inflation?

10 Confirmed increase until Jan 2017

11 Cassandra: Tailored visualisation tools

12 Cassandra: flexibility in input, methods, outputNew sources of data Text mining

13 Cassandra team JRC EUROSTAT Panos Christidis Dario Buono Andris PeizeAlexis Bolivar Álvaro Gomez Losada EUROSTAT Dario Buono Bogomil Kovachev Loredana Di Consiglio Fernando Reis and growing! Contact: